网联环境下基于多驾驶人风险评价的不良行为主动干预研究

依托智能网联技术背景,本文提出基于多驾驶人综合风险评价的不良行为主动干预框架。选取驾驶场景中多驾驶人行为表征参数,基于行车数据定义并辨识驾驶人的各类不良行为;利用面积法实现对不良驾驶行为发生频次、持续时间以及幅值的综合计算,并以可变权重构建其与事故风险的关联关系;借鉴数据包络分析思想,提出考虑可变权重的不良驾驶行为综合评价方法;利用Simulation of Urban Mobility(SUMO)搭建智能网联环境下驾驶仿真平台,以时间窗方式抽取场景中多驾驶人的历史行车数据,利用不良驾驶行为综合评价方法辨识风险驾驶人,提出基于累加窗口与滑动窗口的主动干预方法,每种方法均实现干预单车与干预多车两种策略;分析不同干预手段的效果,探究窗体大小、驾驶人接受率、干预车辆数等对干预结果的影响。研究表明,干预多车策略取得较好的效果,基于累加窗口与滑动窗口的驾驶人不良行为总得分分别下降22.80%和10.50%;与无干预情况相比,当干预接受率为50%时,驾驶人不良行为总得分仍有所降低;与基于累加窗口方法相比,基于滑动窗口的干预方法更适合实际应用。本文提出的框架可为驾驶行为监测等提供技术支持。

Multi-drivers Risk Evaluation Based Proactive Intervention ofDrivers' Risky Behavior Under Connected Transportation Contexts

A proactive intervention framework based on a comprehensive evaluation of risky driving behavior amongmultiple drivers under intelligent and connected transportation contexts is proposed in this study. By selecting drivingbehavior parameters, different types of drivers' risky behavior are defined and identified based on their driving data.Then, an area method is utilized to obtain an integrated score considering the frequency, duration, and amplitude ofeach risky driving behavior, and the relationship between this score and crash risk is established by using variableweights. Based on the mechanism of data envelopment analysis (DEA), a modelling approach for risky drivingbehavior evaluation is proposed. Next, a microsimulation scenario based on Simulation of Urban Mobility(SUMO) isbuilt to simulate the intelligent and connected transportation environment, and the historical driving data of multipledrivers are extracted by a time window. The comprehensive evaluation method of risky driving behavior is applied toidentify risky drivers. Two intervention methods based on a cumulative window approach and a sliding windowapproach are proposed, respectively, and each method implements two strategies, i.e., single-vehicle intervention andmulti-vehicle intervention. Finally, based on the micro-simulation experiment, the effect of different interventionmethods is analyzed, and the impacts of window size, driver acceptance rate, and the number of intervention vehiclesare discussed. The results showed that the strategy of multi-vehicle intervention has achieved better results. The totalscores of drivers' risky behavior under the cumulative window approach and the sliding window method are decreasedby 22.80% and 10.50% , respectively. Compared with the situation without intervention, when the interventionacceptance rate is 50% , the total score of drivers' risky behavior in the scenario still decreases. Relative to thecumulative window approach, the sliding window approach appears to be a more reasonable way for practicalapplication. The proposed framework can provide technical support for driving behavior monitoring.